Error Log Triaging With Explainable Bug Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current error log triaging systems lack trustworthiness and explainability, making it difficult for users to understand and interact with bug reports effectively.
Innovation Solution
A system and method utilizing a classification model to predict a class for a bug report, a recommendation model to identify comparable historical bugs, and an AI model to generate explanation reports, enhancing trust and explainability through consensus reports and a Large Language Model for user interaction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine learning techniques are used for error log triaging, then automation and productivity are improved, but trustworthiness and explainability deteriorate
Solution Approach 1:
The patent introduces explanation reports as an intermediary element between the machine learning system and the user. These reports act as a mediator that translates the black-box ML decisions into human-understandable explanations, thereby maintaining automation while improving trustworthiness. The explanation reports include predicted classes, confidence scores, and key error signatures that bridge the gap between automated processing and user trust.
2Productivity
If machine learning techniques are used for error log triaging, then automation and productivity are improved, but explainability deteriorates
Solution Approach 1:
The patent segments the explainability information into distinct components: predicted class, confidence score, key error signatures, and comparable historical bugs. This segmentation allows the system to maintain automation while providing structured, manageable explanation elements that can be independently processed and presented to users, preventing information overload while preserving explainability.
Solution Approach 2:
Explanation reports serve as an intermediary layer that translates complex machine learning outputs into comprehensible information. The reports include predicted classes with confidence scores and key error signatures that act as intermediaries between the black-box ML model and the user, preserving explainability while maintaining automation.
3Reliability
If manual triaging is used, then trustworthiness and explainability are improved, but productivity deteriorates
Solution Approach 1:
The patent merges the advantages of both manual and automated approaches by combining machine learning automation with explanation report generation. The system automates the initial triaging process while generating explanation reports that provide the trustworthiness and explainability characteristics of manual review, thereby achieving both high productivity and high reliability simultaneously.
4Extent of automation
If black box systems are used, then automation is improved, but ease of operation deteriorates
Solution Approach 1:
The explanation reports act as an intermediary that makes the black-box system operable and understandable for users. By providing predicted classes, confidence scores, and key error signatures in a structured format, the system maintains full automation while dramatically improving ease of operation through enhanced user comprehension and interaction capability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method for error log triaging is described. The method comprises receiving a log file indicative of a bug report and predicting, using a classification model, a class associated with the received log file. The method comprises identifying, using a recommendation model, one or more comparable historical bugs based on the predicted class and a resolution mapping associated with the one or more comparable historical bugs. The method comprises generating, based on an explainability of an Artificial Intelligence (Al) model, a first explanation report associated with the predicted class and a second explanation report associated with the identified one or more comparable historical bugs. The method comprises transmitting, the predicted class, the identified one or more comparable historical bugs, the first explanation report, and the second explanation report for display on a display unit of a user device associated with a triaging manager to facilitate error log triaging.